06:47 <awang> egg|zzz|egg: \o/ downsampling pr

06:47 <awang> Is 228 done after that?

06:47 <awang> Or is there still more to go?

11:14 <egg|work|egg> awang: still needs to be used, but getting close (should be a matter of setting downsampling on the histories of vessels

11:14 <egg|work|egg> )

11:53 <awang> :D

12:15 <awang> Is the change going to increase save file size?

12:15 <awang> Since the entire history is being saved now?

13:08 <egg|work|egg> awang: well the entire history is also being downsampled, so I guess the answer is "it depends"

13:08 <egg|work|egg> where you got #228 artifacts it will have more points, but when doing a 20 min burn which gives you points at 50 Hz, it will massively downsample

13:08 <Qboid> [#228] title: Keep the whole history | For the moment we compute the histories using `sampling_period == 0`, which results in large gaps in the trajectory at high timewarp (rendered as long straight line segments, see below). We should use `sampling_period == 1` instead, and downsample ourselves as needed when rendering.... | https://github.com/mockingbirdnest/Principia/issues/228

13:10 <egg|work|egg> !wa time in new york

13:10 <Qboid> egg|work|egg: current time in New York City, New York, United States: 8:10:51 am EST | Monday, November 6, 2017

13:32 <awang> Oh

13:33 <awang> I thought you were going to keep the entire history and only downsample when rendering

13:41 <egg|work|egg> awang: we do that already

13:42 <egg|work|egg> since Чебышёв

13:43 <egg|work|egg> but in 陈景润 we will also downsample the history at rest when (checking that it's accurate enough)

13:47 <egg|work|egg> s/when //

13:47 <Qboid> egg|work|egg meant to say: but in 陈景润 we will also downsample the history at rest (checking that it's accurate enough)

13:48 <awang> Oh

13:49 <awang> So it should make save files smaller?

13:50 <egg|work|egg> possibly (but it can also make them larger, since we're no longer downsampling them stupidly at high timewarp)

13:53 <awang> Wait

13:53 <awang> So you don't keep the entire history right now?

13:53 <awang> I'm confused

13:54 <egg|work|egg> awang: no, that's why you have #228

13:54 <Qboid> [#228] title: Keep the whole history | For the moment we compute the histories using `sampling_period == 0`, which results in large gaps in the trajectory at high timewarp (rendered as long straight line segments, see below). We should use `sampling_period == 1` instead, and downsample ourselves as needed when rendering.... | https://github.com/mockingbirdnest/Principia/issues/228

13:54 <egg|work|egg> awang: we have at most one point every frame

13:54 <egg|work|egg> so at high warp you can have very sparse points, which causes 228

13:57 <awang> So you keep the entire history of displayed points

13:57 <awang> So what whole history is fixing 228 supposed to let you keep?

13:57 <egg|work|egg> no we don't keep the whole history, that's precisely the problem in 228

13:57 <egg|work|egg> in one frame you may have done 3/4 of an orbit

13:58 <egg|work|egg> (at high timewarp)

13:58 <egg|work|egg> and that's not dense enough to render things correctly

13:58 <egg|work|egg> before Чебышёв it yields an ugly segment, after it yields a silly loop

14:00 <egg|work|egg> 陈景润 will keep the entire history (along those 3/4 of an orbit), but downsampled in a smart way so that it's hopefully reasonably small on disk while remaining accurate

14:07 <GH> [Principia] eggrobin commented on issue #1613: This is part of the work on #228. https://git.io/vFlbd

14:08 <GH> [Principia] eggrobin commented on issue #1617: This is part of the work on #228. https://git.io/vFlbj

14:08 <awang> But you said that you already keep the entire history and only downsample when rendering

14:08 <awang> I'm confused

14:09 <awang> Also, iterating over double?

14:13 <egg|laptop|egg> awang: so, there are two separate things: rendering, and storing

14:14 <egg|laptop|egg> awang: for storing, we currently dowsample to the framerate (with timewarp). This means that you can have points at up to 50 Hz out of timewarp (if engines are lit), and at very high timewarp, points are very sparse, regardless of your trajcetory

14:16 <egg|laptop|egg> awang: for rendering, prior to Чебышёв, we rendered with segments; trajectories looked a bit polygonal, and if the trajectory was stored very densely but looked straight, we were wasting resources drawing more points than needed.

14:17 <egg|laptop|egg> awang: Чебышёв switched to adaptive stepsize rendering of the interpolation, which is much better behaved (smoother, faster) when the information is there

14:18 <egg|laptop|egg> awang: but if the information isn't there, you can't render it; pre-Чебышёв would give you ugly long segments cutting through your trajectory, Чебышёв gives you silly interpolation loops

14:18 <egg|laptop|egg> awang: this is 228

14:19 <egg|laptop|egg> awang: technically, storing the whole history (rather than at most one point per frame) would solve 228

14:19 <egg|laptop|egg> awang: but with one point every 10 s (when engines are not firing), that quickly becomes way too much data if you keep decently long histories

14:20 <egg|laptop|egg> awang: thus we will downsample the storage, but in a way that makes sure our interpolation stays within some tolerance of the original

14:20 <egg|laptop|egg> awang: does that make sense?

14:22 <egg|laptop|egg> awang: as for iterating over double, what's the problem, it's a big superset of int32_t after all

14:25 <egg|laptop|egg> !wpn awang

14:35 <UmbralRaptor> Is that a sword that shatters no matter what?

18:00 <egg|phone|egg> !wpn

18:00 <egg|phone|egg> !Wpn UmbralRaptor

18:00 <egg|phone|egg> !Wpn *

18:00 <UmbralRaptor> !wpn egg|phone|egg

18:34 <egg|afk|egg> !wpn UmbralRaptor

18:35 <UmbralRaptor> *conotoxin

18:39 <egg> UmbralRaptor: are snails cats,

18:40 <UmbralRaptor> uh

20:37 <GH> [Principia] dylanede opened issue #1621: kRPC support https://git.io/vF87f

23:17 <awang> egg|zzz|egg: Ahhh, that makes a lot more sense

23:17 <awang> Thanks for the explanation!

23:18 <awang> And yeah, iterating over double for that particular range works

23:18 <awang> It's just weird